Rows: 6945 Columns: 4
── Column specification ────────────────────────────────────────────────────────
Delimiter: ","
chr (1): ISO2
dbl (3): SALID1, YEAR, deaths
ℹ Use `spec()` to retrieve the full column specification for this data.
ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.
gbmt_log_merged %>%group_by(Clusters, group, YEAR) %>%summarise(deaths_per_100k =mean(deaths_per_100k) ) %>%ungroup() %>%ggplot(aes(x = YEAR, y = deaths_per_100k, color =as.factor(group))) +geom_line(aes(group = group)) +geom_point() +labs(title ="Promedio de Muertes por 100,000 habitantes por Grupo a lo largo del Tiempo",x ="Año", y ="Muertes por 100,000 habitantes", color ="Grupo") +facet_wrap(vars(Clusters)) +theme_minimal()
`summarise()` has grouped output by 'Clusters', 'group'. You can override using
the `.groups` argument.
model_3 <-lmer( deaths_per_100k ~ YEAR * group + (1|SALID1), data = gbmt_log_merged %>%filter(Clusters ==3))summary(model_3)
Linear mixed model fit by REML. t-tests use Satterthwaite's method [
lmerModLmerTest]
Formula: deaths_per_100k ~ YEAR * group + (1 | SALID1)
Data: gbmt_log_merged %>% filter(Clusters == 3)
REML criterion at convergence: 51046.2
Scaled residuals:
Min 1Q Median 3Q Max
-9.2614 -0.3590 -0.0412 0.3379 10.4598
Random effects:
Groups Name Variance Std.Dev.
SALID1 (Intercept) 568422 753.9
Residual 14350 119.8
Number of obs: 3986, groups: SALID1, 253
Fixed effects:
Estimate Std. Error df t value Pr(>|t|)
(Intercept) -15204.6579 1535.9204 3755.7399 -9.899 <2e-16 ***
YEAR 7.9367 0.7637 3730.1916 10.392 <2e-16 ***
group2 -19921.9123 1889.3145 3755.3268 -10.545 <2e-16 ***
group3 -4769.6320 2522.7822 3755.5491 -1.891 0.0588 .
YEAR:group2 10.0842 0.9395 3730.2863 10.734 <2e-16 ***
YEAR:group3 2.4013 1.2545 3730.1344 1.914 0.0557 .
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Correlation of Fixed Effects:
(Intr) YEAR group2 group3 YEAR:2
YEAR -0.998
group2 -0.813 0.812
group3 -0.609 0.608 0.495
YEAR:group2 0.812 -0.813 -0.998 -0.494
YEAR:group3 0.608 -0.609 -0.494 -0.998 0.495
Código
model_4 <-lmer( deaths_per_100k ~ YEAR * group + (1|SALID1), data = gbmt_log_merged %>%filter(Clusters ==4))summary(model_4)
Linear mixed model fit by REML. t-tests use Satterthwaite's method [
lmerModLmerTest]
Formula: deaths_per_100k ~ YEAR * group + (1 | SALID1)
Data: gbmt_log_merged %>% filter(Clusters == 4)
REML criterion at convergence: 51011.2
Scaled residuals:
Min 1Q Median 3Q Max
-9.3443 -0.3480 -0.0255 0.3289 10.5276
Random effects:
Groups Name Variance Std.Dev.
SALID1 (Intercept) 573344 757.2
Residual 14274 119.5
Number of obs: 3986, groups: SALID1, 253
Fixed effects:
Estimate Std. Error df t value Pr(>|t|)
(Intercept) -29967.8679 1235.3518 3754.6485 -24.259 < 2e-16 ***
YEAR 15.4246 0.6143 3730.1321 25.110 < 2e-16 ***
group2 -9544.6955 2079.3163 3754.0162 -4.590 0.00000457 ***
group3 16889.0430 1990.6817 3755.0629 8.484 < 2e-16 ***
group4 11795.1602 3654.4782 3755.8523 3.228 0.001259 **
YEAR:group2 4.7775 1.0340 3729.4454 4.621 0.00000396 ***
YEAR:group3 -8.5523 0.9898 3729.8408 -8.640 < 2e-16 ***
YEAR:group4 -6.0243 1.8172 3728.9163 -3.315 0.000925 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Correlation of Fixed Effects:
(Intr) YEAR group2 group3 group4 YEAR:2 YEAR:3
YEAR -0.998
group2 -0.594 0.593
group3 -0.621 0.620 0.369
group4 -0.338 0.337 0.201 0.210
YEAR:group2 0.593 -0.594 -0.998 -0.368 -0.200
YEAR:group3 0.620 -0.621 -0.368 -0.998 -0.209 0.369
YEAR:group4 0.337 -0.338 -0.200 -0.209 -0.998 0.201 0.210
5.1.2 Poisson
Código
model_3_poisson <-glmer( deaths ~ YEAR * group + (1|SALID1), data = gbmt_log_merged %>%filter(Clusters ==3),family =poisson(link ="log"))
Warning in checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, : Model is nearly unidentifiable: very large eigenvalue
- Rescale variables?;Model is nearly unidentifiable: large eigenvalue ratio
- Rescale variables?
Código
summary(model_3_poisson)
Generalized linear mixed model fit by maximum likelihood (Laplace
Approximation) [glmerMod]
Family: poisson ( log )
Formula: deaths ~ YEAR * group + (1 | SALID1)
Data: gbmt_log_merged %>% filter(Clusters == 3)
AIC BIC logLik deviance df.resid
93071.6 93115.6 -46528.8 93057.6 3979
Scaled residuals:
Min 1Q Median 3Q Max
-27.244 -1.707 -0.105 1.581 34.411
Random effects:
Groups Name Variance Std.Dev.
SALID1 (Intercept) 0.9431 0.9711
Number of obs: 3986, groups: SALID1, 253
Fixed effects:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -28.1982479 0.2120517 -132.98 <2e-16 ***
YEAR 0.0179908 0.0000881 204.21 <2e-16 ***
group2 -3.4199338 0.2662270 -12.85 <2e-16 ***
group3 -15.0532898 0.5040324 -29.87 <2e-16 ***
YEAR:group2 0.0015343 0.0001120 13.70 <2e-16 ***
YEAR:group3 0.0071222 0.0002322 30.67 <2e-16 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Correlation of Fixed Effects:
(Intr) YEAR group2 group3 YEAR:2
YEAR -0.834
group2 -0.797 0.665
group3 -0.420 0.351 0.335
YEAR:group2 0.656 -0.787 -0.845 -0.276
YEAR:group3 0.316 -0.379 -0.252 -0.925 0.298
optimizer (Nelder_Mead) convergence code: 0 (OK)
Model is nearly unidentifiable: very large eigenvalue
- Rescale variables?
Model is nearly unidentifiable: large eigenvalue ratio
- Rescale variables?
Código
model_4_poisson <-glmer( deaths ~ YEAR * group + (1|SALID1), data = gbmt_log_merged %>%filter(Clusters ==4),family =poisson(link ="log"))
Warning in checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, : Model is nearly unidentifiable: very large eigenvalue
- Rescale variables?;Model is nearly unidentifiable: large eigenvalue ratio
- Rescale variables?
Código
summary(model_4_poisson)
Generalized linear mixed model fit by maximum likelihood (Laplace
Approximation) [glmerMod]
Family: poisson ( log )
Formula: deaths ~ YEAR * group + (1 | SALID1)
Data: gbmt_log_merged %>% filter(Clusters == 4)
AIC BIC logLik deviance df.resid
91102.0 91158.7 -45542.0 91084.0 3977
Scaled residuals:
Min 1Q Median 3Q Max
-27.499 -1.706 -0.088 1.531 34.979
Random effects:
Groups Name Variance Std.Dev.
SALID1 (Intercept) 0.9275 0.9631
Number of obs: 3986, groups: SALID1, 253
Fixed effects:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -28.74899723 0.17419126 -165.043 <2e-16 ***
YEAR 0.01814074 0.00007399 245.190 <2e-16 ***
group2 -17.82188573 0.37318440 -47.756 <2e-16 ***
group3 0.12327807 0.27597423 0.447 0.655
group4 -10.49954899 0.77767851 -13.501 <2e-16 ***
YEAR:group2 0.00865560 0.00016947 51.075 <2e-16 ***
YEAR:group3 0.00006760 0.00011577 0.584 0.559
YEAR:group4 0.00497635 0.00036088 13.790 <2e-16 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Correlation of Fixed Effects:
(Intr) YEAR group2 group3 group4 YEAR:2 YEAR:3
YEAR -0.853
group2 -0.467 0.398
group3 -0.631 0.538 0.295
group4 -0.224 0.191 0.105 0.141
YEAR:group2 0.372 -0.436 -0.912 -0.235 -0.083
YEAR:group3 0.545 -0.639 -0.254 -0.842 -0.122 0.279
YEAR:group4 0.175 -0.205 -0.082 -0.110 -0.932 0.089 0.131
optimizer (Nelder_Mead) convergence code: 0 (OK)
Model is nearly unidentifiable: very large eigenvalue
- Rescale variables?
Model is nearly unidentifiable: large eigenvalue ratio
- Rescale variables?
5.2 Modelo 2 - Con offset
5.2.1 Estimación
Estimación de los modelos con la información poblacional como offset
Código
model_3_poisson_off <-glmer( deaths ~ YEAR * group + (1|SALID1) +offset(log(population_imp)), data = gbmt_log_merged %>%filter(Clusters ==3),family =poisson(link ="log"))
Warning in checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, :
Model failed to converge with max|grad| = 0.00333122 (tol = 0.002, component 1)
Warning in checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, : Model is nearly unidentifiable: very large eigenvalue
- Rescale variables?;Model is nearly unidentifiable: large eigenvalue ratio
- Rescale variables?
Código
summary(model_3_poisson_off)
Generalized linear mixed model fit by maximum likelihood (Laplace
Approximation) [glmerMod]
Family: poisson ( log )
Formula: deaths ~ YEAR * group + (1 | SALID1) + offset(log(population_imp))
Data: gbmt_log_merged %>% filter(Clusters == 3)
AIC BIC logLik deviance df.resid
125757.1 125801.2 -62871.6 125743.1 3979
Scaled residuals:
Min 1Q Median 3Q Max
-27.043 -1.978 -0.110 1.933 35.567
Random effects:
Groups Name Variance Std.Dev.
SALID1 (Intercept) 0.2631 0.5129
Number of obs: 3986, groups: SALID1, 253
Fixed effects:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 5.65891920 0.18770436 30.15 <2e-16 ***
YEAR -0.00536626 0.00008828 -60.79 <2e-16 ***
group2 -43.12373070 0.23740987 -181.64 <2e-16 ***
group3 -36.89556012 0.47936169 -76.97 <2e-16 ***
YEAR:group2 0.02167489 0.00011215 193.27 <2e-16 ***
YEAR:group3 0.01844462 0.00023332 79.05 <2e-16 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Correlation of Fixed Effects:
(Intr) YEAR group2 group3 YEAR:2
YEAR -0.944
group2 -0.791 0.747
group3 -0.393 0.371 0.310
YEAR:group2 0.744 -0.787 -0.949 -0.292
YEAR:group3 0.358 -0.379 -0.283 -0.977 0.299
optimizer (Nelder_Mead) convergence code: 0 (OK)
Model failed to converge with max|grad| = 0.00333122 (tol = 0.002, component 1)
Model is nearly unidentifiable: very large eigenvalue
- Rescale variables?
Model is nearly unidentifiable: large eigenvalue ratio
- Rescale variables?
Código
model_4_poisson_off <-glmer( deaths ~ YEAR * group + (1|SALID1) +offset(log(population_imp)), data = gbmt_log_merged %>%filter(Clusters ==4),family =poisson(link ="log"))
Warning in checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, : Model is nearly unidentifiable: very large eigenvalue
- Rescale variables?;Model is nearly unidentifiable: large eigenvalue ratio
- Rescale variables?
Código
summary(model_4_poisson_off)
Generalized linear mixed model fit by maximum likelihood (Laplace
Approximation) [glmerMod]
Family: poisson ( log )
Formula: deaths ~ YEAR * group + (1 | SALID1) + offset(log(population_imp))
Data: gbmt_log_merged %>% filter(Clusters == 4)
AIC BIC logLik deviance df.resid
119296.0 119352.6 -59639.0 119278.0 3977
Scaled residuals:
Min 1Q Median 3Q Max
-25.965 -1.993 -0.072 1.944 36.208
Random effects:
Groups Name Variance Std.Dev.
SALID1 (Intercept) 0.2627 0.5126
Number of obs: 3986, groups: SALID1, 253
Fixed effects:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -33.91616080 0.15652727 -216.679 <2e-16 ***
YEAR 0.01453835 0.00007411 196.170 <2e-16 ***
group2 -19.42859056 0.35019735 -55.479 <2e-16 ***
group3 41.17822747 0.24616549 167.279 <2e-16 ***
group4 7.19853240 0.74096565 9.715 <2e-16 ***
YEAR:group2 0.00963692 0.00016957 56.832 <2e-16 ***
YEAR:group3 -0.02071338 0.00011605 -178.481 <2e-16 ***
YEAR:group4 -0.00374071 0.00036137 -10.352 <2e-16 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Correlation of Fixed Effects:
(Intr) YEAR group2 group3 group4 YEAR:2 YEAR:3
YEAR -0.951
group2 -0.448 0.426
group3 -0.636 0.605 0.285
group4 -0.211 0.201 0.095 0.135
YEAR:group2 0.416 -0.438 -0.972 -0.265 -0.088
YEAR:group3 0.608 -0.639 -0.272 -0.947 -0.129 0.280
YEAR:group4 0.195 -0.205 -0.087 -0.124 -0.979 0.090 0.131
optimizer (Nelder_Mead) convergence code: 0 (OK)
Model is nearly unidentifiable: very large eigenvalue
- Rescale variables?
Model is nearly unidentifiable: large eigenvalue ratio
- Rescale variables?
Ingreso de los valores predictivos para los modelos de 3 y 4 clústers:
Debido a que se ha agregado a los interceptos de las ciudades como efecto aleatorio, entonces las estimaciones y predicciones van a variar de ciudad a ciudad. Para poder visualizarlo, se podría resumir los datos con su media.
Código
plot_poisson_offset <- gbmt_log_merged %>%group_by(Clusters, group, YEAR) %>%summarise(across(c(predicted_deaths_100k, deaths_per_100k), mean ) ) %>%ungroup() %>%ggplot(aes(x = YEAR, y = predicted_deaths_100k,color = group)) +geom_point(aes(y = deaths_per_100k), alpha =0.5,position =position_jitter(h =0.2)) +# Datos observadosgeom_line() +labs(title ="Promedio de Muertes por 100,000 habitantes por Grupo",x ="Año", y ="Muertes esperadas por 100,000 habitantes", color ="Grupo") +facet_wrap(vars(Clusters)) +theme_minimal()
`summarise()` has grouped output by 'Clusters', 'group'. You can override using
the `.groups` argument.
Las estimaciones no consideran variaciones entre las ciudades
Código
model_3_poisson_off2 <-glm( deaths ~ YEAR * group +offset(log(population_imp)), data = gbmt_log_merged %>%filter(Clusters ==3),family =poisson(link ="log"))summary(model_3_poisson_off2)
Call:
glm(formula = deaths ~ YEAR * group + offset(log(population_imp)),
family = poisson(link = "log"), data = gbmt_log_merged %>%
filter(Clusters == 3))
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 20.91471063 0.17356342 120.5 <2e-16 ***
YEAR -0.01298050 0.00008645 -150.2 <2e-16 ***
group2 -83.35453870 0.21854660 -381.4 <2e-16 ***
group3 -58.38416556 0.45938267 -127.1 <2e-16 ***
YEAR:group2 0.04187171 0.00010884 384.7 <2e-16 ***
YEAR:group3 0.02918818 0.00022878 127.6 <2e-16 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
(Dispersion parameter for poisson family taken to be 1)
Null deviance: 7403371 on 3985 degrees of freedom
Residual deviance: 5079058 on 3980 degrees of freedom
AIC: 5116575
Number of Fisher Scoring iterations: 5
Código
model_4_poisson_off2 <-glm( deaths ~ YEAR * group +offset(log(population_imp)), data = gbmt_log_merged %>%filter(Clusters ==4),family =poisson(link ="log"))summary(model_4_poisson_off2)
Call:
glm(formula = deaths ~ YEAR * group + offset(log(population_imp)),
family = poisson(link = "log"), data = gbmt_log_merged %>%
filter(Clusters == 4))
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -61.03060884 0.14169673 -430.71 <2e-16 ***
YEAR 0.02818326 0.00007055 399.46 <2e-16 ***
group2 -4.04165649 0.33191778 -12.18 <2e-16 ***
group3 83.26129484 0.22552208 369.19 <2e-16 ***
group4 33.79128295 0.71482476 47.27 <2e-16 ***
YEAR:group2 0.00195728 0.00016528 11.84 <2e-16 ***
YEAR:group3 -0.04182671 0.00011231 -372.41 <2e-16 ***
YEAR:group4 -0.01704939 0.00035601 -47.89 <2e-16 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
(Dispersion parameter for poisson family taken to be 1)
Null deviance: 7403371 on 3985 degrees of freedom
Residual deviance: 5149021 on 3978 degrees of freedom
AIC: 5186542
Number of Fisher Scoring iterations: 5
gbmt_log_merged2 %>%group_by(Clusters, group, YEAR) %>%summarise(across(c(predicted_deaths_100k, deaths_per_100k), mean ) ) %>%ungroup() %>%ggplot(aes(x = YEAR, y = predicted_deaths_100k,color = group)) +geom_point(aes(y = deaths_per_100k), alpha =0.5,position =position_jitter(h =0.2)) +# Datos observadosgeom_line(linewidth =1) +labs(title ="Promedio de Muertes por 100,000 habitantes por Grupo a lo largo del Tiempo",x ="Año", y ="Muertes esperadas por 100,000 habitantes", color ="Grupo") +facet_wrap(vars(Clusters)) +theme_minimal()
`summarise()` has grouped output by 'Clusters', 'group'. You can override using
the `.groups` argument.
Sin resumir:
Código
gbmt_log_merged2 %>%ggplot(aes(x = YEAR, y = predicted_deaths_100k,color = group)) +geom_point(aes(y = deaths_per_100k), alpha =0.5,position =position_jitter(h =0.2)) +# Datos observadosgeom_line(linewidth =1) +labs(title ="Promedio de Muertes por 100,000 habitantes por Grupo a lo largo del Tiempo",x ="Año", y ="Muertes esperadas por 100,000 habitantes", color ="Grupo") +facet_wrap(vars(Clusters)) +theme_minimal()
